case CHIP_VEGA20: chip_string = "AMD RADV VEGA20"; break;
case CHIP_RAVEN: chip_string = "AMD RADV RAVEN"; break;
case CHIP_RAVEN2: chip_string = "AMD RADV RAVEN2"; break;
+ case CHIP_NAVI10: chip_string = "AMD RADV NAVI10"; break;
+ case CHIP_NAVI12: chip_string = "AMD RADV NAVI12"; break;
+ case CHIP_NAVI14: chip_string = "AMD RADV NAVI14"; break;
default: chip_string = "AMD RADV unknown"; break;
}
/* Override family and chip_class. */
device->rad_info.family = i;
- if (i >= CHIP_VEGA10)
+ if (i >= CHIP_NAVI10)
+ device->rad_info.chip_class = GFX10;
+ else if (i >= CHIP_VEGA10)
device->rad_info.chip_class = GFX9;
else if (i >= CHIP_TONGA)
device->rad_info.chip_class = GFX8;
/* The mere presence of CLEAR_STATE in the IB causes random GPU hangs
* on GFX6.
*/
- device->has_clear_state = device->rad_info.chip_class >= GFX7;
+ device->has_clear_state = device->rad_info.chip_class >= GFX7 &&
+ device->rad_info.chip_class <= GFX9;
device->cpdma_prefetch_writes_memory = device->rad_info.chip_class <= GFX8;
(device->rad_info.chip_class >= GFX8 &&
device->rad_info.me_fw_feature >= 41);
- device->has_dcc_constant_encode = device->rad_info.family == CHIP_RAVEN2;
+ device->has_dcc_constant_encode = device->rad_info.family == CHIP_RAVEN2 ||
+ device->rad_info.chip_class >= GFX10;
device->use_shader_ballot = device->instance->perftest_flags & RADV_PERFTEST_SHADER_BALLOT;
.fullDrawIndexUint32 = true,
.imageCubeArray = true,
.independentBlend = true,
- .geometryShader = true,
- .tessellationShader = true,
+ .geometryShader = pdevice->rad_info.chip_class < GFX10,
+ .tessellationShader = pdevice->rad_info.chip_class < GFX10,
.sampleRateShading = true,
.dualSrcBlend = true,
.logicOp = true,
}
}
- device->pbb_allowed = device->physical_device->rad_info.chip_class >= GFX9 &&
+ /* TODO: Enable binning for GFX10. */
+ device->pbb_allowed = device->physical_device->rad_info.chip_class == GFX9 &&
!(device->instance->debug_flags & RADV_DEBUG_NOBINNING);
/* Disabled and not implemented for now. */
S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
- S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
- S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
S_008F0C_ELEMENT_SIZE(1) |
S_008F0C_INDEX_STRIDE(3) |
S_008F0C_ADD_TID_ENABLE(true);
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ desc[3] |= S_008F0C_FORMAT(V_008F0C_IMG_FORMAT_32_FLOAT) |
+ S_008F0C_OOB_SELECT(2) |
+ S_008F0C_RESOURCE_LEVEL(1);
+ } else {
+ desc[3] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
+ }
+
/* GS entry for ES->GS ring */
/* stride 0, num records - size, elsize0,
index stride 0 */
S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
- S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
- S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
S_008F0C_ELEMENT_SIZE(0) |
S_008F0C_INDEX_STRIDE(0) |
S_008F0C_ADD_TID_ENABLE(false);
+
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ desc[7] |= S_008F0C_FORMAT(V_008F0C_IMG_FORMAT_32_FLOAT) |
+ S_008F0C_OOB_SELECT(2) |
+ S_008F0C_RESOURCE_LEVEL(1);
+ } else {
+ desc[7] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
+ }
}
desc += 8;
S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
- S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
- S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
S_008F0C_ELEMENT_SIZE(0) |
S_008F0C_INDEX_STRIDE(0) |
S_008F0C_ADD_TID_ENABLE(false);
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ desc[3] |= S_008F0C_FORMAT(V_008F0C_IMG_FORMAT_32_FLOAT) |
+ S_008F0C_OOB_SELECT(2) |
+ S_008F0C_RESOURCE_LEVEL(1);
+ } else {
+ desc[3] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
+ }
+
/* stride gsvs_itemsize, num records 64
elsize 4, index stride 16 */
/* shader will patch stride and desc[2] */
S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
- S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
- S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
S_008F0C_ELEMENT_SIZE(1) |
S_008F0C_INDEX_STRIDE(1) |
S_008F0C_ADD_TID_ENABLE(true);
+
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ desc[7] |= S_008F0C_FORMAT(V_008F0C_IMG_FORMAT_32_FLOAT) |
+ S_008F0C_OOB_SELECT(2) |
+ S_008F0C_RESOURCE_LEVEL(1);
+ } else {
+ desc[7] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
+ }
+
}
desc += 8;
desc[3] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X) |
S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
- S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
- S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
- S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
- S_008F0C_ELEMENT_SIZE(0) |
- S_008F0C_INDEX_STRIDE(0) |
- S_008F0C_ADD_TID_ENABLE(false);
+ S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W);
+
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ desc[3] |= S_008F0C_FORMAT(V_008F0C_IMG_FORMAT_32_FLOAT) |
+ S_008F0C_OOB_SELECT(3) |
+ S_008F0C_RESOURCE_LEVEL(1);
+ } else {
+ desc[3] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
+ }
desc[4] = tess_offchip_va;
desc[5] = S_008F04_BASE_ADDRESS_HI(tess_offchip_va >> 32) |
desc[7] = S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X) |
S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
- S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
- S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
- S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32) |
- S_008F0C_ELEMENT_SIZE(0) |
- S_008F0C_INDEX_STRIDE(0) |
- S_008F0C_ADD_TID_ENABLE(false);
+ S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W);
+
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ desc[7] |= S_008F0C_FORMAT(V_008F0C_IMG_FORMAT_32_FLOAT) |
+ S_008F0C_OOB_SELECT(3) |
+ S_008F0C_RESOURCE_LEVEL(1);
+ } else {
+ desc[7] |= S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32);
+ }
}
desc += 8;
S_030938_SIZE(tf_ring_size / 4));
radeon_set_uconfig_reg(cs, R_030940_VGT_TF_MEMORY_BASE,
tf_va >> 8);
- if (queue->device->physical_device->rad_info.chip_class >= GFX9) {
+
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ radeon_set_uconfig_reg(cs, R_030984_VGT_TF_MEMORY_BASE_HI_UMD,
+ S_030984_BASE_HI(tf_va >> 40));
+ } else if (queue->device->physical_device->rad_info.chip_class == GFX9) {
radeon_set_uconfig_reg(cs, R_030944_VGT_TF_MEMORY_BASE_HI,
S_030944_BASE_HI(tf_va >> 40));
}
radv_cs_add_buffer(queue->device->ws, cs, descriptor_bo);
- if (queue->device->physical_device->rad_info.chip_class >= GFX9) {
+ if (queue->device->physical_device->rad_info.chip_class >= GFX10) {
+ uint32_t regs[] = {R_00B030_SPI_SHADER_USER_DATA_PS_0,
+ R_00B130_SPI_SHADER_USER_DATA_VS_0,
+ R_00B230_SPI_SHADER_USER_DATA_GS_0,
+ R_00B430_SPI_SHADER_USER_DATA_HS_0};
+
+ for (int i = 0; i < ARRAY_SIZE(regs); ++i) {
+ radv_emit_shader_pointer(queue->device, cs, regs[i],
+ va, true);
+ }
+ } else if (queue->device->physical_device->rad_info.chip_class >= GFX9) {
uint32_t regs[] = {R_00B030_SPI_SHADER_USER_DATA_PS_0,
R_00B130_SPI_SHADER_USER_DATA_VS_0,
R_00B208_SPI_SHADER_USER_DATA_ADDR_LO_GS,
device->physical_device->rad_info.chip_class <= GFX8)
va += plane->surface.u.legacy.level[iview->base_mip].dcc_offset;
+ unsigned dcc_tile_swizzle = surf->tile_swizzle;
+ dcc_tile_swizzle &= (surf->dcc_alignment - 1) >> 8;
+
cb->cb_dcc_base = va >> 8;
- cb->cb_dcc_base |= surf->tile_swizzle;
+ cb->cb_dcc_base |= dcc_tile_swizzle;
/* GFX10 field has the same base shift as the GFX6 field. */
uint32_t max_slice = radv_surface_max_layer_count(iview) - 1;
uint32_t max_slice = radv_surface_max_layer_count(iview) - 1;
ds->db_depth_view = S_028008_SLICE_START(iview->base_layer) |
S_028008_SLICE_MAX(max_slice);
+ if (device->physical_device->rad_info.chip_class >= GFX10) {
+ ds->db_depth_view |= S_028008_SLICE_START_HI(iview->base_layer >> 11) |
+ S_028008_SLICE_MAX_HI(max_slice >> 11);
+ }
ds->db_htile_data_base = 0;
ds->db_htile_surface = 0;
ds->db_stencil_info = S_02803C_FORMAT(stencil_format) |
S_02803C_SW_MODE(surf->u.gfx9.stencil.swizzle_mode);
- ds->db_z_info2 = S_028068_EPITCH(surf->u.gfx9.surf.epitch);
- ds->db_stencil_info2 = S_02806C_EPITCH(surf->u.gfx9.stencil.epitch);
- ds->db_depth_view |= S_028008_MIPID(level);
+ if (device->physical_device->rad_info.chip_class == GFX9) {
+ ds->db_z_info2 = S_028068_EPITCH(surf->u.gfx9.surf.epitch);
+ ds->db_stencil_info2 = S_02806C_EPITCH(surf->u.gfx9.stencil.epitch);
+ }
+ ds->db_depth_view |= S_028008_MIPID(level);
ds->db_depth_size = S_02801C_X_MAX(iview->image->info.width - 1) |
S_02801C_Y_MAX(iview->image->info.height - 1);
unsigned max_zplanes =
radv_calc_decompress_on_z_planes(device, iview);
- ds->db_z_info |= S_028038_DECOMPRESS_ON_N_ZPLANES(max_zplanes) |
- S_028038_ITERATE_FLUSH(1);
- ds->db_stencil_info |= S_02803C_ITERATE_FLUSH(1);
+ ds->db_z_info |= S_028038_DECOMPRESS_ON_N_ZPLANES(max_zplanes);
+
+ if (device->physical_device->rad_info.chip_class >= GFX10) {
+ ds->db_z_info |= S_028040_ITERATE_FLUSH(1);
+ ds->db_stencil_info |= S_028044_ITERATE_FLUSH(1);
+ } else {
+ ds->db_z_info |= S_028038_ITERATE_FLUSH(1);
+ ds->db_stencil_info |= S_02803C_ITERATE_FLUSH(1);
+ }
}
if (!surf->has_stencil)
iview->image->htile_offset;
ds->db_htile_data_base = va >> 8;
ds->db_htile_surface = S_028ABC_FULL_CACHE(1) |
- S_028ABC_PIPE_ALIGNED(surf->u.gfx9.htile.pipe_aligned) |
- S_028ABC_RB_ALIGNED(surf->u.gfx9.htile.rb_aligned);
+ S_028ABC_PIPE_ALIGNED(surf->u.gfx9.htile.pipe_aligned);
+
+ if (device->physical_device->rad_info.chip_class == GFX9) {
+ ds->db_htile_surface |= S_028ABC_RB_ALIGNED(surf->u.gfx9.htile.rb_aligned);
+ }
}
} else {
const struct legacy_surf_level *level_info = &surf->u.legacy.level[level];
{
uint32_t max_aniso = radv_get_max_anisotropy(device, pCreateInfo);
uint32_t max_aniso_ratio = radv_tex_aniso_filter(max_aniso);
- bool is_vi = (device->physical_device->rad_info.chip_class >= GFX8);
+ bool compat_mode = device->physical_device->rad_info.chip_class == GFX8 ||
+ device->physical_device->rad_info.chip_class == GFX9;
unsigned filter_mode = V_008F30_SQ_IMG_FILTER_MODE_BLEND;
const struct VkSamplerReductionModeCreateInfoEXT *sampler_reduction =
S_008F30_ANISO_THRESHOLD(max_aniso_ratio >> 1) |
S_008F30_ANISO_BIAS(max_aniso_ratio) |
S_008F30_DISABLE_CUBE_WRAP(0) |
- S_008F30_COMPAT_MODE(is_vi) |
+ S_008F30_COMPAT_MODE(compat_mode) |
S_008F30_FILTER_MODE(filter_mode));
sampler->state[1] = (S_008F34_MIN_LOD(S_FIXED(CLAMP(pCreateInfo->minLod, 0, 15), 8)) |
S_008F34_MAX_LOD(S_FIXED(CLAMP(pCreateInfo->maxLod, 0, 15), 8)) |
S_008F38_XY_MAG_FILTER(radv_tex_filter(pCreateInfo->magFilter, max_aniso)) |
S_008F38_XY_MIN_FILTER(radv_tex_filter(pCreateInfo->minFilter, max_aniso)) |
S_008F38_MIP_FILTER(radv_tex_mipfilter(pCreateInfo->mipmapMode)) |
- S_008F38_MIP_POINT_PRECLAMP(0) |
- S_008F38_DISABLE_LSB_CEIL(device->physical_device->rad_info.chip_class <= GFX8) |
- S_008F38_FILTER_PREC_FIX(1) |
- S_008F38_ANISO_OVERRIDE_GFX6(is_vi));
+ S_008F38_MIP_POINT_PRECLAMP(0));
sampler->state[3] = (S_008F3C_BORDER_COLOR_PTR(0) |
S_008F3C_BORDER_COLOR_TYPE(radv_tex_bordercolor(pCreateInfo->borderColor)));
+
+ if (device->physical_device->rad_info.chip_class >= GFX10) {
+ sampler->state[2] |= S_008F38_ANISO_OVERRIDE_GFX10(1);
+ } else {
+ sampler->state[2] |=
+ S_008F38_DISABLE_LSB_CEIL(device->physical_device->rad_info.chip_class <= GFX8) |
+ S_008F38_FILTER_PREC_FIX(1) |
+ S_008F38_ANISO_OVERRIDE_GFX6(device->physical_device->rad_info.chip_class >= GFX8);
+ }
}
VkResult radv_CreateSampler(
projects / mesa.git / blobdiff